Primary sclerosing cholangitis (PSC) is a rare, broad clinical spectrum, immune-mediated disease of the biliary tract that slowly progresses toward cirrhosis and biliary complications. The heterogeneity and poor prevalence of PSC, its uncertain pathophysiology, its slowly progression toward clinical endpoints, and the lack of surrogate endpoints are the major challenges in risk stratification and drug discovery. In fact, despite numerous trials performed, no therapies have been currently registered for the disease. Magnetic resonance cholangiopancreatography (MRI-MRCP) is fundamental for diagnosing, staging, and monitoring PSC progression and, therefore, it might be a promising candidate as a surrogate biomarker of disease progression. However, prognostic scores routinely used in clinical practice do not include imaging-derived biomarkers. This is mainly due to their qualitative/semi-quantitative nature, which compromises the inter- and intra-observer agreement. The general aim of this Ph.D. project is to identify novel quantitative imaging-based biomarkers for disease monitoring and drug response evaluation in PSC using innovative image analysis techniques based on the application of artificial intelligence(AI). The specific aims of this project are: - to identify whether quantitative biliary metrics, obtained from routinely collected MRCP images using an AI-derived software(MRCP+®), are predictive of clinical outcomes of PSC patients; - to identify if radiomics features, extracted from conventional MRI-MRCP images acquired with a standardized acquisition protocol, can classify PSC patients at higher risk of disease progression; - to test the derived imaging-based biomarkers in a randomized-controlled clinical trial as exploratory endpoints for the evaluation of treatment benefit. Using both retrospective and prospective data of MRI-MRCP images acquired from PSC patients: - we derived a predictive model, the quantitative MRCP-PSC score, including two imaging-based biomarkers: the number of strictures derived using MRCP+® and the spleen length. Adjusted for age, sex, and time from diagnosis to MRCP, these biomarkers were able to predict the development of hepatobiliary complications with a C-Statistic of 0.80. After internal validation, our score outperformed the already available prognostic radiological score in PSC (the Anali score without gadolinium) with a C-statistic of 0.78 vs 0.64; - we identified five radiomics features, extracted using a semi-automated process of segmentation and free and universal software for feature extraction (PyRadiomics). These features were able to identify patients at higher risk of progression using the available estimates of outcome (Mayo risk score and Liver Stiffness Measure) showing a mean AUCs close to or higher than 80%, which reached 96% in the multivariable analysis; - we designed a phase 2 randomized placebo-controlled clinical trial to assess the safety and efficacy of oral vancomycin in patients with PSC, including as exploratory endpoint the changes in the imaging-based biomarkers derived in this thesis’ studies. The results of the studies included in this thesis confirm the promising role of imaging-based biomarkers in risk stratification in PSC. In addition, the upcoming clinical trial will also assess their potential role in evaluating the benefits following a promising treatment by monitoring their changes in a shorter time horizon, such as that of a clinical trial.

La colangite sclerosante primitiva(CSP) è una malattia rara, immuno-mediata delle vie biliari che può progredire verso la cirrosi epatica e complicanze biliari. Ad oggi, non esistono ancora terapie registrate per il trattamento della malattia e l’unica soluzione è il trapianto epatico. La presentazione clinica molto eterogenea, la sua scarsa prevalenza e l’incerta patogenesi contribuiscono all’assenza di biomarcatori surrogati di andamento di malattia e di valutazione di risposta al trattamento che compromettono la scoperta di nuove terapie. L’imaging radiologico, in particolare la risonanza magnetica con studio colangiografico(MRI-MRCP), è fondamentale per la diagnosi, la stadiazione e il monitoraggio della progressione della CSP. Tuttavia, gli score prognostici utilizzati nella pratica clinica non includono variabili basati sulle immagini di risonanza. Ciò è dovuto alla loro natura qualitativa/semi-quantitativa, che compromette l'accordo inter- e intra-osservatore. L'obiettivo principale di questo progetto di dottorato consiste nell’identificare nuovi biomarcatori quantitativi radiologici per il monitoraggio della malattia e la valutazione della risposta alla terapia nella CSP utilizzando tecniche innovative di analisi delle immagini basate sull'applicazione dell'intelligenza artificiale (AI). Gli obiettivi specifici di questo progetto sono: - identificare metriche biliari quantitative, ottenute dall’analisi delle sequenze colangiografiche di risonanza magnetica con l’utilizzo di un software sviluppato con l’applicazione dall'AI(MRCP +®), predittive di outcome clinici nei pazienti con CSP; - identificare variabili derivate con l’uso della radiomica da MRI-MRCP convenzionali acquisite con un protocollo di acquisizione standardizzato che possano classificare i pazienti CSP a più alto rischio di progressione della malattia utilizzando le stime di outcome attualmente disponibili nella pratica clinica. - testare i biomarcatori radiologici derivati in questa tesi in uno studio clinico per valutarne le prestazioni nella valutazione della risposta a una determinata terapia. Utilizzando dati retrospettivi e prospettici dei pazienti con CSP seguiti presso la Fondazione IRCCS San Gerardo dei Tintori: - abbiamo derivato un modello predittivo, il “quantitative MRCP-PSC score”, che include due biomarcatori radiologici: il numero di stenosi derivato dall’analisi di MRCP+® e la lunghezza della milza. Aggiustati per età, sesso e tempo dalla diagnosi alla MRI-MRCP, questi biomarcatori sono in grado di prevedere lo sviluppo di complicanze epatobiliari con una C-statistic di 0,80. Dopo la validazione interna, il nostro score ha dimostrato una migliore performance rispetto allo score radiologico prognostico disponibile in CSP con una C-statistic di 0,78 vs 0,64; - abbiamo identificato cinque features radiomiche, estratte utilizzando un processo semi-automatico di segmentazione e un software gratuito e universale di estrazione (PyRadiomics). Queste caratteristiche sono in grado di identificare i pazienti a più alto rischio di progressione utilizzando le stime di esito disponibili (Mayo risk score ed elastografia epatica) mostrando un'area sotto la curva (AUC) media vicina o superiore all'80%, che ha raggiunto il 96% nell'analisi multivariata; - abbiamo progettato uno studio clinico di fase 2, randomizzato controllato con placebo, per valutare la sicurezza e l'efficacia della vancomicina orale in pazienti con PSC, includendo come endpoint esplorativi i cambiamenti nei biomarcatori basati sull'imaging derivati negli studi di questa tesi. I risultati di questa tesi confermano il ruolo promettente dei biomarcatori basati sull'imaging nella stratificazione del rischio nella PSC. Inoltre, la prossima sperimentazione clinica valuterà anche il loro potenziale ruolo nella valutazione dei benefici a seguito di un trattamento monitorando i loro cambiamenti in un orizzonte temporale più breve.

(2023). RADIOLOGY-BASED BIOMARKERS DISCOVERY IN PRIMARY SCLEROSING CHOLANGITIS: FROM RISK STRATIFICATION TO THE EVALUATION OF TREATMENT BENEFIT. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).

RADIOLOGY-BASED BIOMARKERS DISCOVERY IN PRIMARY SCLEROSING CHOLANGITIS: FROM RISK STRATIFICATION TO THE EVALUATION OF TREATMENT BENEFIT

CRISTOFERI, LAURA
2023

Abstract

Primary sclerosing cholangitis (PSC) is a rare, broad clinical spectrum, immune-mediated disease of the biliary tract that slowly progresses toward cirrhosis and biliary complications. The heterogeneity and poor prevalence of PSC, its uncertain pathophysiology, its slowly progression toward clinical endpoints, and the lack of surrogate endpoints are the major challenges in risk stratification and drug discovery. In fact, despite numerous trials performed, no therapies have been currently registered for the disease. Magnetic resonance cholangiopancreatography (MRI-MRCP) is fundamental for diagnosing, staging, and monitoring PSC progression and, therefore, it might be a promising candidate as a surrogate biomarker of disease progression. However, prognostic scores routinely used in clinical practice do not include imaging-derived biomarkers. This is mainly due to their qualitative/semi-quantitative nature, which compromises the inter- and intra-observer agreement. The general aim of this Ph.D. project is to identify novel quantitative imaging-based biomarkers for disease monitoring and drug response evaluation in PSC using innovative image analysis techniques based on the application of artificial intelligence(AI). The specific aims of this project are: - to identify whether quantitative biliary metrics, obtained from routinely collected MRCP images using an AI-derived software(MRCP+®), are predictive of clinical outcomes of PSC patients; - to identify if radiomics features, extracted from conventional MRI-MRCP images acquired with a standardized acquisition protocol, can classify PSC patients at higher risk of disease progression; - to test the derived imaging-based biomarkers in a randomized-controlled clinical trial as exploratory endpoints for the evaluation of treatment benefit. Using both retrospective and prospective data of MRI-MRCP images acquired from PSC patients: - we derived a predictive model, the quantitative MRCP-PSC score, including two imaging-based biomarkers: the number of strictures derived using MRCP+® and the spleen length. Adjusted for age, sex, and time from diagnosis to MRCP, these biomarkers were able to predict the development of hepatobiliary complications with a C-Statistic of 0.80. After internal validation, our score outperformed the already available prognostic radiological score in PSC (the Anali score without gadolinium) with a C-statistic of 0.78 vs 0.64; - we identified five radiomics features, extracted using a semi-automated process of segmentation and free and universal software for feature extraction (PyRadiomics). These features were able to identify patients at higher risk of progression using the available estimates of outcome (Mayo risk score and Liver Stiffness Measure) showing a mean AUCs close to or higher than 80%, which reached 96% in the multivariable analysis; - we designed a phase 2 randomized placebo-controlled clinical trial to assess the safety and efficacy of oral vancomycin in patients with PSC, including as exploratory endpoint the changes in the imaging-based biomarkers derived in this thesis’ studies. The results of the studies included in this thesis confirm the promising role of imaging-based biomarkers in risk stratification in PSC. In addition, the upcoming clinical trial will also assess their potential role in evaluating the benefits following a promising treatment by monitoring their changes in a shorter time horizon, such as that of a clinical trial.
VALSECCHI, MARIA GRAZIA
CARBONE, MARCO
BERNASCONI, DAVIDE PAOLO
Radiomica; Colangite; Malattia rara; Biomarcatori; Radiologia
Radiomics; Rare disease; Cholangitis; Biomarkers; Prognostic
MED/01 - STATISTICA MEDICA
Italian
20-apr-2023
35
2021/2022
embargoed_20260420
(2023). RADIOLOGY-BASED BIOMARKERS DISCOVERY IN PRIMARY SCLEROSING CHOLANGITIS: FROM RISK STRATIFICATION TO THE EVALUATION OF TREATMENT BENEFIT. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_854333.pdf

embargo fino al 20/04/2026

Descrizione: Tesi
Tipologia di allegato: Doctoral thesis
Dimensione 10.72 MB
Formato Adobe PDF
10.72 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/416558
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact